The Frappe MCP server exhibits significant quality gaps across naming, descriptions, parameter documentation, and output schema clarity. While tool names follow verb-noun conventions (get_, list_, create_, update_, delete_, search_), descriptions are sparse and many are generic. Critical issues: (1) Parameter descriptions are minimal or absent, most tools lack field-level documentation beyond basic type info. (2) Output schemas are not documented anywhere in the source code. (3) Three tools (calculate_project_metrics, project_risk_assessment, portfolio_dashboard) are marked as 'fabricated - not dispatchable' in test files, indicating incomplete implementation. (4) Error handling is not visible, no recovery guidance, error categorization, or actionable failure messages. (5) Many domain-specific tools (analyze_document, analyze_project_timeline, generate_project_report) have vague descriptions that do not explain when to use them or what they return. The server defines 16 tools total, but 3 appear non-functional and the remaining 13 have inconsistent quality. Average tool score: 38/100.

• Remove or complete the three fabricated tools (calculate_project_metrics, project_risk_assessment, portfolio_dashboard) marked as non-dispatchable.
• Document output schemas for all 13 dispatchable tools. Include field names, data types, required vs optional fields, and brief field descriptions. Use a schema doc format (JSON Schema or OpenAPI) visible in the tool definition.
• Expand parameter descriptions to 50-150 characters each. For 'data' params, list expected fields (e.g., 'Object with keys: subject (string), body (string), priority (enum: low|medium|high)). For 'filters', document the filter syntax (e.g., 'MongoDB-style or SQL WHERE clause syntax').
• Add error handling guidance to each tool description. Document common failure modes and recovery steps: 'If document not found, try search_documents() with a partial name. If permission denied, check your Frappe role.'
• Rewrite vague domain-specific tool descriptions. For analyze_document: 'AI-powered analysis of document content. Returns sentiment, key entities, and summary. Use before generating reports or when user asks to understand a document's core themes.'
• For list_documents, clarify pagination: 'Returns up to 'limit' documents (default 20, max 100). To fetch more, use offset parameter (not shown, add if not present) or implement cursor-based pagination.' If limit is unbounded, cap it at a safe value (e.g., 50).
• Add permission scope annotations to tool descriptions. E.g., delete_document: '[Requires: doctype:delete permission] Permanently deletes a document. This action is irreversible.'

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry